Results of the present study confirm that ovarian response to a standard dose of FSH mainly depends on ovarian reserve. Indeed, in the present study the number of retrieved oocytes is significantly predicted by age and two markers of ovarian reserve namely AFC and FSH, with the last two variables being the most significant predictors.
The multiple follicular growth in cycles of assisted reproduction is effected using exogenous FSH, leading to supranormal circulating concentrations and recruitment of all follicles whose FSH sensitivity threshold is exceeded [14, 15]. When exogenous FSH is administered, the number of follicles induced to grow largely depends upon the number of follicles and their FSH sensitivity [16, 17]. That is the main reason why an algorithm including AFC and serum d3FSH may correctly predict the number of retrieved oocytes following ovarian stimulation.
The prediction of oocyte yield is the basis for tailoring the treatment in the daily clinical practice. Indeed the choice of the appropriate starting dose of FSH can only be secondary to an accurate prediction of oocyte yield following maximal ovarian stimulation. If the predicted oocyte yield is suggestive for excessive ovarian response the starting dose of FSH should be lower than the one leading to the maximal stimulation. For this reason, some authors examined and tested complex models based on multiple phenotypic, ultrasound derived and biochemical indexes to dictate starting doses of exogenous gonadotrophins in IVF cycles [18, 19].
A first prospective study showed that the combination of age, antral follicle count (AFC), ovarian volume, Doppler ovarian score and smoking status may allow clinicians to choose the appropriate FSH dose in IVF cycles . In a second study the proposed model was based on age, body mass index (BMI), serum day 3 FSH (d3FSH) and AFC . Both models were then validated in successive prospective trials, demonstrating that the application of an individualized versus standard FSH dose was associated with a reduced cancellation rate for abnormal ovarian response, reduced need of adjusting the dose during treatment and increased occurrence of an adequate ovarian response [20, 21]. Unfortunately both the published nomograms may not be used by clinicians in their practice since the nomogram by Popovic-Todorovic  was based on a Doppler score of ovarian stromal blood flow and testosterone levels, which are not commonly measured in the clinical practice, whereas the model created by Howles  has never been publicly disclosed.
In the present study ovarian response is significantly predicted by age, serum FSH and AFC and this model may permit the construction of an FSH dosage nomogram. The nomogram (Figure 1) was constructed after setting the desirable number of retrieved oocytes as nine. As already said, the hypothesis was that stimulating the ovaries wishing to obtain a number of oocytes close to the median in the whole population leads to a narrow distribution of ovarian responses around that value. This would imply that fewer women would exhibit an inadequate response (i.e. poor or excessive response). This methodology has been successfully used in previously published models [13, 18, 19]. Moreover nine is the middle point of the appropriate ovarian response when defined as the retrieval of 5 to 14 oocytes [19, 22] and the middlepoint between the cut-off values usually used to define “poor response” (<4 oocytes)  and “excessive ovarian response” (>15) [8, 24–26].
The need of individualizing doses of FSH in patients derives from the assumption that variability in the functional ovarian reserve (the pool of recruitable follicles) is very wide; therefore a standard fixed dose of FSH may not be suitable for all women . Indeed several observations [19, 21, 22] indicate that 150 IU of FSH is the optimal dose for part of the studied patients, with others requiring higher doses of FSH in order to reach the optimal stimulation. On the other hand a large proportion of OHSS cases are preventable by reducing the FSH starting dose in women at risk of excessive ovarian response, namely those with a large pool of recruitable follicles.
Everything previously enounced clearly indicates that the starting dose of FSH in IVF cycles should be individualized. In the present study we clearly demonstrated that the FSH daily dose may be calculated on the basis of the age of patients and of two markers of ovarian reserve, namely AFC and FSH, with the last two variables being the most significant predictors. The model based on AFC resembles a previously published model based on AMH . The two models have the same accuracy in predicting ovarian response to gonadotropins and this is not surprising since a linear strong relationship exists between AFC and AMH [4, 7]. The strong correlation between the two markers reflects the fact that the same ovarian follicles which are seen on ultrasound secrete AMH. One relevant benefit of AFC is the possibility of its measurement at the same moment in which clinicians examine the patient hence we developed a second simplified model based only on age and AFC (Figure 2). As expected the two variables model is less accurate than the one incorporating age, FSH and AFC, but being independent of the blood sample makes it highly interesting for clinicians.
When the nomogram we developed was tested in the same population used to elaborate the model (Figure 3), it predicted a dose lower than 150 IU in 8% of patients, a dose between 150 and 187.5 IU in 11.5% of patients, or between 187.5 and 225 IU in 10.5% of patients. In women ≤ 35 years the calculated FSH starting dose was lower than 225 IU in 50.2% of patients while it was lower than 225 IU only in 18.1% of women > 35 years. Before clinicians can adopt the model into routine clinical practice, the accuracy of the model should be independently evaluated in a population different from the one on which the model was elaborated. External validation is therefore crucial to assess the generalizability of our model to other populations.